Why Are Fewer Adults Surpassing Their Parents’ Incomes?
I began reading the new Raj Chetty paper—what Richard Reeves calls the “new Chetty bomb”—not wanting to believe it, because it shows a large drop in absolute social mobility. But I also suspected that it might be right.
Partly, that’s because the Chetty team has done such great work on mobility previously. But I also knew that income growth unambiguously has slowed since the 1960s. That’s very different from saying that incomes have declined, which they’ve not. But the rate at which our living standards have improved has diminished. If incomes used to rise faster, then it follows that past generations were more likely to out-do their parents than today’s adults.
As I read through the paper, Chetty and his authors uncannily anticipated my potential criticisms, one by one. I don’t think there is any real doubt that their main finding is correct: many fewer adults exceed their parents’ income today than in in the past.
Why do Chetty’s findings differ from those of previous researchers?
One initial puzzle was how their results could be consistent with past work on absolute mobility using longitudinal household surveys. The Pew Economic Mobility Project, which once employed me as research manager, found that 67 percent of today’s (roughly) 40-year-olds are better off than their parents were, which is significantly higher than the 50 percent absolute mobility rate reported for today’s (roughly) 30-year-olds by Chetty et al. After adjusting incomes for the number of household members that share them, Pew found the absolute mobility rate to be even higher — 84 percent. Both of the Pew studies used a dataset called the Panel Study of Income Dynamics (PSID). Using another survey, the National Longitudinal Survey of Youth 1979 (NLSY79), I found the size-adjusted rate to be 83 percent (in an essay that, I should admit, looks not-so-great in the wake of the Chetty paper).
These estimates have the virtue of being based on actual observation of adult incomes and their childhood incomes. How can the Chetty results, which for earlier decades are based on fancy modeling, be right if these estimates are correct?
Older research, showing higher mobility, came before the decline was complete
Well, it turns out that the PSID and NLSY79 estimates came from cohorts that were smack-dab in the middle of the long-term absolute mobility decline. The Pew research focused on adults who were zero to 18 years old in 1968, meaning they were born between 1950 and 1968. That captures a lot of kids in the cohorts where absolute mobility hadn’t stabilized at a new low level (see Figure 1B of the Chetty paper). In 1950, 80 percent of 30-year-olds saw upward absolute mobility, compared with 60 percent in 1968 and 50 percent in 1980. The adult children in the later Pew study were also a few years older, on average, than their parents were when their incomes were measured, which pushes absolute mobility up a bit too.
My estimates using the NLSY79 examined adults who were born in 1962, 1963, and 1964. Like the Pew research, I tried to measure incomes around age 40, while the Chetty paper focuses on 30-year-old incomes. His paper’s Figure 3D indicates an absolute mobility rate (using size-adjusted incomes of 30-year-olds) of about 70 percent for the 1962 through 1964 birth cohorts, and from Figure 3C, it looks like using 40-year-old incomes would bump that up by around 5 points. Using the “PCE deflator” to adjust incomes for the increase in the cost of living (as I did) instead of their chosen price index would bump it up a tiny bit more (see their Figure 3A).
Call it a 76 percent rate of absolute mobility. That’s lower by a few percentage points than my estimate, but my estimate is from imperfect data too. Regardless, I don’t see how a 76 percent rate is higher than whatever the corresponding 1940 rate would be, and it’s clearly well above the range of rates Chetty shows for the 1984 cohort.
(Side note: Chetty’s estimates come from assuming that relative mobility — the extent to which adult and parent income ranks are related — was the same for cohorts born before the 1980s as it is was for cohorts born in the early 1980s. The fact that their 1962–64 estimate lines up reasonably well with what the NLSY79 shows is more evidence that relative mobility was no worse for children born in the early 1980s as it was for children born in the early 1960s.)
Two beefs with the Chetty bomb
So I’m pretty confident the trend in the Chetty paper is right. I have two beefs with the paper, however. First, because it is about the trend, the authors are less focused on getting the level of absolute mobility right in any given year. That’s unfortunate because much of the media coverage focused on the paper’s finding that just half of today’s 30-year-olds are doing better than their parents. That’s almost certainly an underestimate.
The best estimates of the level of absolute mobility in any year would be derived from different measurement choices than those in the headline numbers from the report. It isn’t possible to produce these best estimates, especially in earlier decades, but it is safe to say that they would show more absolute mobility than Chetty’s headline figures.
For instance, Figure 3D in their paper shows estimates that use size-adjusted income, producing a 60 percent absolute mobility rate for the 1984 cohort instead of a 50 percent rate. Figure 3A shows that using the PCE deflator to adjust incomes for the cost of living increases the absolute mobility rate by 3 or 4 percentage points. Figure 3B shows that accounting for government transfers and taxes raises the rate by two points.
Finally, Figure S4 shows that if we assume that the incomes of everyone not experiencing absolute mobility in the baseline numbers actually are higher by $5,000 than the baseline figures indicate, that would push up the share of the 1984 cohort achieving upward mobility by about 5 percentage points. Why might that be a reasonable assumption? Health benefits and other nonwage compensation are one factor. Nothing in the Chetty paper includes such benefits as income. Cohabitation is another. Two cohabiters will be two “families” in the Census Bureau data used in the paper (and will be two “tax units” in the paper’s tax data). In reality, cohabiters pool their incomes, just like married couples.
Undercounting of income is a third reason to think that the reported incomes in the Chetty paper are too low. Undercounting is a pretty bad problem in the bottom third, especially in the CPS and census data, but also potentially in the tax data, where people don’t report under-the-table earnings. (I reviewed this evidence in Appendix 3 of my recent paper on poverty trends). A caveat here is that parents also have understated incomes because of these issues, but nonwage compensation, government health benefits, cohabitation, and undercounting of income have all increased over time, so their impact is greater on children.
Put all this together and it looks to me like size-adjusting pushes the absolute mobility rate up 10 points, using the PCE deflator another 3 to 4 points, taxes and transfers another 2, and the rest (plausibly) another 5 points. That’s 20–24 percentage points, which would put the absolute mobility rate at 70–74 percent. Two-thirds seems like a safe conservative estimate.
For adults who were poor children, absolute mobility rates almost certainly remain above 90 percent. This is hardly evidence that the American Dream is “fading,” as the paper’s title claims. The period from 1939 to 1969 was one of exceptionally strong income growth. That growth translated into very high absolute mobility rates. Income growth has slowed since then, though it has not reversed. Thus, absolute mobility rates have fallen, though most people still do better than their parents did at the same age.
My second beef is with their conclusion, as stated in the abstract, that:
Increasing GDP growth rates alone cannot restore absolute mobility to the rates experienced by children born in the 1940s. In contrast, changing the distribution of growth across income groups to the more equal distribution experienced by the 1940 birth cohort would reverse more than 70% of the decline in mobility. These results imply that reviving the “American Dream” of high rates of absolute mobility would require economic growth that is spread more broadly across the income distribution.
Lower economic growth = lower inequality
There is an issue here in that economic growth might have been even weaker if inequality had declined. This is a big question that is nowhere close to being answered. Elsewhere, I have summarized the mess that is the research literature on this question, and while that review is already a bit out of date, the basic conclusions seem solid to me. There is a fair amount of (weak) evidence that rising income concentration increases growth rates in developed English-speaking nations. The slowdown in income growth began about a decade before income concentration started to take off, which raises doubt about the extent to which inequality caused the slower growth. And since 1980, anyway, income growth below the top has been fastest during periods when income concentration has risen.
It is hard to think of a scenario in which we might have achieved equal growth across the income distribution in recent decades (let alone a decline in inequality). One can imagine that a smaller share of income growth might have gone to the top, but because of globalization and increasing demand for skill (and rising inequality in the educational distribution), the relevant counterfactual to the past generation or two probably isn’t that the bottom 90 percentiles would have grown by the same amount, it’s that, say, percentiles 75 to 90 would have enjoyed much more of the “freed-up” growth than the bottom half (to the extent that slowed economic growth didn’t whittle away that “freed-up” amount).
Chetty and his coauthors succumb to inequality alarmism when they say, “higher GDP growth itself cannot increase absolute mobility unless it is more broadly distributed.” Their own analyses show that if we’d had the growth rates of the 1940, 1950s, and 1960s while inequality grew as it did, absolute mobility for the 1980 birth cohort would have been 62 percent instead of 50 percent. And as noted above, these levels are underestimates of absolute mobility levels. A 12-point increase in absolute mobility from stronger growth would raise a 65 percent true mobility rate to 77 percent.
Would it be preferable to have the even higher absolute mobility rate Chetty and his coauthors claim we’d have if inequality had declined? Maybe — a lot depends on why inequality rose and on how we’d have prevented that from happening. But knowing that rising inequality reduces absolute mobility tells us nothing about the causes of rising inequality, nor about the effects — on absolute mobility and otherwise — of different policies that might have reversed it or might in the future.
Americans are as rich as they’ve ever been
Absolute mobility remains robust, especially for poor kids. Americans are as rich as they’ve ever been. Faster economic growth would improve income growth and absolute mobility without obvious risks to the economy. Reducing inequality might increase absolute mobility too. In particular, there may be growth-promoting policies that reduce inequality by, for instance, reducing crony capitalism or expanding competition. Those policies are worth pursuing because they would promote growth, not because they would reduce inequality per se.
Most of all, what we need is a conversation nuanced enough to recognize that there are tradeoffs to achieving perfect mobility rates and there are some social and economic changes that we cannot alter via policy. I have often argued that relative mobility doesn’t have to be lower than it used to be for us to think it is too low. It is just as true that absolute mobility doesn’t have to be as high as it used to be to think it is still reasonably healthy. But we can and should aspire to more of both relative and absolute mobility.